1 paper
Yijiong Yu, Shuai Yuan, Jie Zheng +2
Soft context compression reduces the computational workload of processing long contexts in LLMs by encoding long context into a smaller number of latent tokens. However, existing f…